Move task_complete command out of prompt (Significant-Gravitas#3663)

* feat: move task_complete command out of prompt * fix: formatting fixes * Add the shutdown command to the test agents * tests: update test vcrs --------- Co-authored-by: James Collins <[email protected]>
oralf1011 · May 3, 2023 · d744280 · d744280
1 parent e21917c
commit d744280
Show file tree

Hide file tree

Showing 9 changed files with 1,415 additions and 27 deletions.
diff --git a/autogpt/app.py b/autogpt/app.py
@@ -120,8 +120,6 @@ def execute_command(
         # TODO: Change these to take in a file rather than pasted code, if
         # non-file is given, return instructions "Input should be a python
         # filepath, write your code to file and try again
-        elif command_name == "task_complete":
-            shutdown()
         else:
             for command in prompt.commands:
                 if (
@@ -171,12 +169,6 @@ def get_hyperlinks(url: str) -> Union[str, List[str]]:
     return scrape_links(url)
 
 
-def shutdown() -> NoReturn:
-    """Shut down the program"""
-    logger.info("Shutting down...")
-    quit()
-
-
 @command(
     "start_agent",
     "Start GPT Agent",

diff --git a/autogpt/commands/task_statuses.py b/autogpt/commands/task_statuses.py
@@ -0,0 +1,26 @@
+"""Task Statuses module."""
+from __future__ import annotations
+
+from typing import NoReturn
+
+from autogpt.commands.command import command
+from autogpt.logs import logger
+
+
+@command(
+    "task_complete",
+    "Task Complete (Shutdown)",
+    '"reason": "<reason>"',
+)
+def task_complete(reason: str) -> NoReturn:
+    """
+    A function that takes in a string and exits the program
+
+    Parameters:
+        reason (str): The reason for shutting down.
+    Returns:
+        A result string from create chat completion. A list of suggestions to
+            improve the code.
+    """
+    logger.info(title="Shutting down...\n", message=reason)
+    quit()
diff --git a/autogpt/main.py b/autogpt/main.py
@@ -130,6 +130,7 @@ def run_auto_gpt(
     command_registry.import_commands("autogpt.commands.web_selenium")
     command_registry.import_commands("autogpt.commands.write_tests")
     command_registry.import_commands("autogpt.app")
+    command_registry.import_commands("autogpt.commands.task_statuses")
 
     ai_name = ""
     ai_config = construct_main_ai_config()

diff --git a/autogpt/prompts/prompt.py b/autogpt/prompts/prompt.py
@@ -41,15 +41,6 @@ def build_default_prompt_generator() -> PromptGenerator:
         'Exclusively use the commands listed in double quotes e.g. "command name"'
     )
 
-    # Define the command list
-    commands = [
-        ("Task Complete (Shutdown)", "task_complete", {"reason": "<reason>"}),
-    ]
-
-    # Add commands to the PromptGenerator object
-    for command_label, command_name, args in commands:
-        prompt_generator.add_command(command_label, command_name, args)
-
     # Add resources to the PromptGenerator object
     prompt_generator.add_resource(
         "Internet access for searches and information gathering."

diff --git a/tests/integration/agent_factory.py b/tests/integration/agent_factory.py
@@ -45,6 +45,7 @@ def browser_agent(agent_test_config, memory_none: NoMemory, workspace: Workspace
     command_registry.import_commands("autogpt.commands.file_operations")
     command_registry.import_commands("autogpt.commands.web_selenium")
     command_registry.import_commands("autogpt.app")
+    command_registry.import_commands("autogpt.commands.task_statuses")
 
     ai_config = AIConfig(
         ai_name="browse_website-GPT",
@@ -80,6 +81,7 @@ def writer_agent(agent_test_config, memory_none: NoMemory, workspace: Workspace)
     command_registry = CommandRegistry()
     command_registry.import_commands("autogpt.commands.file_operations")
     command_registry.import_commands("autogpt.app")
+    command_registry.import_commands("autogpt.commands.task_statuses")
 
     ai_config = AIConfig(
         ai_name="write_to_file-GPT",
@@ -120,6 +122,7 @@ def memory_management_agent(
     command_registry = CommandRegistry()
     command_registry.import_commands("autogpt.commands.file_operations")
     command_registry.import_commands("autogpt.app")
+    command_registry.import_commands("autogpt.commands.task_statuses")
 
     ai_config = AIConfig(
         ai_name="Follow-Instructions-GPT",

diff --git a/...egration/challenges/memory/cassettes/test_memory_challenge_a/test_memory_challenge_a.yaml b/...egration/challenges/memory/cassettes/test_memory_challenge_a/test_memory_challenge_a.yaml
diff --git a/tests/integration/goal_oriented/cassettes/test_browse_website/test_browse_website.yaml b/tests/integration/goal_oriented/cassettes/test_browse_website/test_browse_website.yaml
diff --git a/tests/integration/goal_oriented/cassettes/test_write_file/test_write_file.yaml b/tests/integration/goal_oriented/cassettes/test_write_file/test_write_file.yaml
@@ -324,4 +324,253 @@ interactions:
     status:
       code: 200
       message: OK
+- request:
+    body: '{"model": "gpt-3.5-turbo", "messages": [{"role": "system", "content": "You
+      are write_to_file-GPT, an AI designed to use the write_to_file command to write
+      ''Hello World'' into a file named \"hello_world.txt\" and then use the task_complete
+      command to complete the task.\nYour decisions must always be made independently
+      without seeking user assistance. Play to your strengths as an LLM and pursue
+      simple strategies with no legal complications.\n\nGOALS:\n\n1. Use the write_to_file
+      command to write ''Hello World'' into a file named \"hello_world.txt\".\n2.
+      Use the task_complete command to complete the task.\n3. Do not use any other
+      commands.\n\n\nConstraints:\n1. ~4000 word limit for short term memory. Your
+      short term memory is short, so immediately save important information to files.\n2.
+      If you are unsure how you previously did something or want to recall past events,
+      thinking about similar events will help you remember.\n3. No user assistance\n4.
+      Exclusively use the commands listed in double quotes e.g. \"command name\"\n\nCommands:\n1.
+      append_to_file: Append to file, args: \"filename\": \"<filename>\", \"text\":
+      \"<text>\"\n2. delete_file: Delete file, args: \"filename\": \"<filename>\"\n3.
+      list_files: List Files in Directory, args: \"directory\": \"<directory>\"\n4.
+      read_file: Read file, args: \"filename\": \"<filename>\"\n5. write_to_file:
+      Write to file, args: \"filename\": \"<filename>\", \"text\": \"<text>\"\n6.
+      delete_agent: Delete GPT Agent, args: \"key\": \"<key>\"\n7. get_hyperlinks:
+      Get text summary, args: \"url\": \"<url>\"\n8. get_text_summary: Get text summary,
+      args: \"url\": \"<url>\", \"question\": \"<question>\"\n9. list_agents: List
+      GPT Agents, args: () -> str\n10. message_agent: Message GPT Agent, args: \"key\":
+      \"<key>\", \"message\": \"<message>\"\n11. start_agent: Start GPT Agent, args:
+      \"name\": \"<name>\", \"task\": \"<short_task_desc>\", \"prompt\": \"<prompt>\"\n12.
+      task_complete: Task Complete (Shutdown), args: \"reason\": \"<reason>\"\n\nResources:\n1.
+      Internet access for searches and information gathering.\n2. Long Term memory
+      management.\n3. GPT-3.5 powered Agents for delegation of simple tasks.\n4. File
+      output.\n\nPerformance Evaluation:\n1. Continuously review and analyze your
+      actions to ensure you are performing to the best of your abilities.\n2. Constructively
+      self-criticize your big-picture behavior constantly.\n3. Reflect on past decisions
+      and strategies to refine your approach.\n4. Every command has a cost, so be
+      smart and efficient. Aim to complete tasks in the least number of steps.\n5.
+      Write all code to a file.\n\nYou should only respond in JSON format as described
+      below \nResponse Format: \n{\n    \"thoughts\": {\n        \"text\": \"thought\",\n        \"reasoning\":
+      \"reasoning\",\n        \"plan\": \"- short bulleted\\n- list that conveys\\n-
+      long-term plan\",\n        \"criticism\": \"constructive self-criticism\",\n        \"speak\":
+      \"thoughts summary to say to user\"\n    },\n    \"command\": {\n        \"name\":
+      \"command name\",\n        \"args\": {\n            \"arg name\": \"value\"\n        }\n    }\n}
+      \nEnsure the response can be parsed by Python json.loads"}, {"role": "system",
+      "content": "The current time and date is Tue Jan  1 00:00:00 2000"}, {"role":
+      "user", "content": "Determine which next command to use, and respond using the
+      format specified above:"}], "temperature": 0, "max_tokens": 0}'
+    headers:
+      Accept:
+      - '*/*'
+      Accept-Encoding:
+      - gzip, deflate
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '3406'
+      Content-Type:
+      - application/json
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAA7yTT4/TQAzF73wKy5de0iqh0O3mthwQPYAEdMUuBFWzE7cZOhmHGUctVPnuKEn/
+        7AZxQnD1c9772eMc0OSYoi6U6LKy46tXP+5vbpZhoT+/u329nCc/39692H58PzUf7q8xQn74RlqO
+        X0w0l5UlMewwQu1JCeWYJrP5NJ7Pr5J5hCXnZDHFTSXj6eTlWGr/wON4GicYYR3UhjA9YOW5rGQl
+        vCUXML2aJRFevM/153EcobAoey5dz5ImQl2w0RQw/XLAksLJ1rMlTFGFYIIoJy0kOyHXDnDIHABA
+        hlJwvSkkZJjCsXgUaC9tMcMFOKIchKEOBFIQ7LwRWgmv1sYSaC5L5bqGToDRG7KW4RN7m4/AOGFQ
+        0LU6VVIOo6LVV7tWn8heRpMMo8fZnlRgZ9ymB1gWBKLCFjx9r42nACX9RVoEHe0f5zChE1VVea68
+        UXKR1uxBirZBhe2QurLK9cBjuP1Pm9LeiNEmlMOnIhdq3zIogcX54TR7T1q6iDah3wTt5YIk3HW2
+        HcOwUJHanoJ2xtp/eRB9cBOdDvVo+tudtg490xOIAbrym+GJ98JpE73FgOOJyfC3eDTTibYjPoJn
+        rsEmwrVxJhSr/qIxxSBcYYTG5bTHNG6+Ns9+AQAA//8DAAaMC/aGBAAA
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      CF-RAY:
+      - 7c15d0c86db14665-DFW
+      Cache-Control:
+      - no-cache, must-revalidate
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 May 2023 04:38:50 GMT
+      Server:
+      - cloudflare
+      access-control-allow-origin:
+      - '*'
+      alt-svc:
+      - h3=":443"; ma=86400, h3-29=":443"; ma=86400
+      openai-model:
+      - gpt-3.5-turbo-0301
+      openai-organization:
+      - significant-gravitas
+      openai-processing-ms:
+      - '11864'
+      openai-version:
+      - '2020-10-01'
+      strict-transport-security:
+      - max-age=15724800; includeSubDomains
+      x-ratelimit-limit-requests:
+      - '3500'
+      x-ratelimit-limit-tokens:
+      - '90000'
+      x-ratelimit-remaining-requests:
+      - '3499'
+      x-ratelimit-remaining-tokens:
+      - '86496'
+      x-ratelimit-reset-requests:
+      - 17ms
+      x-ratelimit-reset-tokens:
+      - 2.335s
+      x-request-id:
+      - 5d9f1c3b077997ea564ea7088eca733f
+    status:
+      code: 200
+      message: OK
+- request:
+    body: '{"model": "gpt-3.5-turbo", "messages": [{"role": "system", "content": "You
+      are write_to_file-GPT, an AI designed to use the write_to_file command to write
+      ''Hello World'' into a file named \"hello_world.txt\" and then use the task_complete
+      command to complete the task.\nYour decisions must always be made independently
+      without seeking user assistance. Play to your strengths as an LLM and pursue
+      simple strategies with no legal complications.\n\nGOALS:\n\n1. Use the write_to_file
+      command to write ''Hello World'' into a file named \"hello_world.txt\".\n2.
+      Use the task_complete command to complete the task.\n3. Do not use any other
+      commands.\n\n\nConstraints:\n1. ~4000 word limit for short term memory. Your
+      short term memory is short, so immediately save important information to files.\n2.
+      If you are unsure how you previously did something or want to recall past events,
+      thinking about similar events will help you remember.\n3. No user assistance\n4.
+      Exclusively use the commands listed in double quotes e.g. \"command name\"\n\nCommands:\n1.
+      append_to_file: Append to file, args: \"filename\": \"<filename>\", \"text\":
+      \"<text>\"\n2. delete_file: Delete file, args: \"filename\": \"<filename>\"\n3.
+      list_files: List Files in Directory, args: \"directory\": \"<directory>\"\n4.
+      read_file: Read file, args: \"filename\": \"<filename>\"\n5. write_to_file:
+      Write to file, args: \"filename\": \"<filename>\", \"text\": \"<text>\"\n6.
+      delete_agent: Delete GPT Agent, args: \"key\": \"<key>\"\n7. get_hyperlinks:
+      Get text summary, args: \"url\": \"<url>\"\n8. get_text_summary: Get text summary,
+      args: \"url\": \"<url>\", \"question\": \"<question>\"\n9. list_agents: List
+      GPT Agents, args: () -> str\n10. message_agent: Message GPT Agent, args: \"key\":
+      \"<key>\", \"message\": \"<message>\"\n11. start_agent: Start GPT Agent, args:
+      \"name\": \"<name>\", \"task\": \"<short_task_desc>\", \"prompt\": \"<prompt>\"\n12.
+      task_complete: Task Complete (Shutdown), args: \"reason\": \"<reason>\"\n\nResources:\n1.
+      Internet access for searches and information gathering.\n2. Long Term memory
+      management.\n3. GPT-3.5 powered Agents for delegation of simple tasks.\n4. File
+      output.\n\nPerformance Evaluation:\n1. Continuously review and analyze your
+      actions to ensure you are performing to the best of your abilities.\n2. Constructively
+      self-criticize your big-picture behavior constantly.\n3. Reflect on past decisions
+      and strategies to refine your approach.\n4. Every command has a cost, so be
+      smart and efficient. Aim to complete tasks in the least number of steps.\n5.
+      Write all code to a file.\n\nYou should only respond in JSON format as described
+      below \nResponse Format: \n{\n    \"thoughts\": {\n        \"text\": \"thought\",\n        \"reasoning\":
+      \"reasoning\",\n        \"plan\": \"- short bulleted\\n- list that conveys\\n-
+      long-term plan\",\n        \"criticism\": \"constructive self-criticism\",\n        \"speak\":
+      \"thoughts summary to say to user\"\n    },\n    \"command\": {\n        \"name\":
+      \"command name\",\n        \"args\": {\n            \"arg name\": \"value\"\n        }\n    }\n}
+      \nEnsure the response can be parsed by Python json.loads"}, {"role": "system",
+      "content": "The current time and date is Tue Jan  1 00:00:00 2000"}, {"role":
+      "system", "content": "This reminds you of these events from your past: \nI was
+      created and nothing new has happened."}, {"role": "user", "content": "Determine
+      which next command to use, and respond using the format specified above:"},
+      {"role": "assistant", "content": "{\n    \"thoughts\": {\n        \"text\":
+      \"I need to use the write_to_file command to write ''Hello World'' into a file
+      named ''hello_world.txt''.\",\n        \"reasoning\": \"The task requires me
+      to write ''Hello World'' into a file named ''hello_world.txt'', and the write_to_file
+      command is the appropriate command for this task.\",\n        \"plan\": \"-
+      Use the write_to_file command to write ''Hello World'' into a file named ''hello_world.txt''.\",\n        \"criticism\":
+      \"I need to ensure that I use the correct filename and text to write to the
+      file.\",\n        \"speak\": \"I will use the write_to_file command to write
+      ''Hello World'' into a file named ''hello_world.txt''.\"\n    },\n    \"command\":
+      {\n        \"name\": \"write_to_file\",\n        \"args\": {\n            \"filename\":
+      \"hello_world.txt\",\n            \"text\": \"Hello World\"\n        }\n    }\n}"},
+      {"role": "system", "content": "Command write_to_file returned: File written
+      to successfully."}, {"role": "user", "content": "Determine which next command
+      to use, and respond using the format specified above:"}], "temperature": 0,
+      "max_tokens": 0}'
+    headers:
+      Accept:
+      - '*/*'
+      Accept-Encoding:
+      - gzip, deflate
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '4653'
+      Content-Type:
+      - application/json
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    body:
+      string: !!binary |
+        H4sIAAAAAAAAA7RTyW7bMBC99ysGc/GFNmQbbVwdiwapgQLd0hZBVBgMNZZYU6RKjrzA0L8XtKQk
+        TReg242YefMWcnhEnWOKqpSsqtqMz54drl5V+Z6W/OaDfHlhzpMNmefnh4vX5i0KdDefSXE/MVGu
+        qg2xdhYFKk+SKcd0+mQxTxaLs/lMYOVyMphiUfN4Pnk85sbfuHEyT6YosAmyIEyPWHtX1bxityEb
+        MJ0ms4XAO/LbxiyZCmTH0txhZ7OnrUBVOq0oYHp9xIrCQOydIUxRhqADS8vRprNMNkY4ZhYAIEMu
+        XVOUHDJMoS/2DdpzLGa4hFJuCUKjFIWwbow5wM5rZrIwekHGOPjovMlHoC07kLDWhsDKinIYlbG/
+        2sX+hPc8msBlSWBpzxCYatAB2EETCLgkYBk2qz48gXJVJW0eAbe1ATXJUNy360kGZ7UtOs+XPQw8
+        fWm0pwAVRZ7om37fNZxslGR/YATeaasIfnVL7E7wqCBOp3+fvzbSdtHH8P6v2ZTXrJUO1bABlug0
+        STY0Ps5JhiXU3m11TiBhK43OoXsEWDs/SGhb/FQk1CQ3g8BOG/Nn19BRtmLY6B7+3ULHp+3UvqF/
+        YEr64uFfuL9g/+NHDBFOMfo0mW2xFbjWVody1WljioFdjQK1zWmPadJ+ah99BQAA//8DAMo52K3G
+        BAAA
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      CF-RAY:
+      - 7c15d120ec8f4665-DFW
+      Cache-Control:
+      - no-cache, must-revalidate
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 May 2023 04:39:04 GMT
+      Server:
+      - cloudflare
+      access-control-allow-origin:
+      - '*'
+      alt-svc:
+      - h3=":443"; ma=86400, h3-29=":443"; ma=86400
+      openai-model:
+      - gpt-3.5-turbo-0301
+      openai-organization:
+      - significant-gravitas
+      openai-processing-ms:
+      - '12271'
+      openai-version:
+      - '2020-10-01'
+      strict-transport-security:
+      - max-age=15724800; includeSubDomains
+      x-ratelimit-limit-requests:
+      - '3500'
+      x-ratelimit-limit-tokens:
+      - '90000'
+      x-ratelimit-remaining-requests:
+      - '3499'
+      x-ratelimit-remaining-tokens:
+      - '86481'
+      x-ratelimit-reset-requests:
+      - 17ms
+      x-ratelimit-reset-tokens:
+      - 2.346s
+      x-request-id:
+      - a8c02e37e5af0c851759aa261f20f701
+    status:
+      code: 200
+      message: OK
 version: 1
diff --git a/tests/integration/goal_oriented/test_browse_website.py b/tests/integration/goal_oriented/test_browse_website.py
@@ -11,7 +11,7 @@
 def test_browse_website(browser_agent: Agent) -> None:
     file_path = browser_agent.workspace.get_path("browse_website.txt")
     try:
-        run_interaction_loop(browser_agent, 40)
+        run_interaction_loop(browser_agent, 120)
     # catch system exit exceptions
     except SystemExit:  # the agent returns an exception when it shuts down
         content = read_file(file_path)